This is the seventh week of my GSoC journey.

GSoC project

• GSoC project URL
• Work Repository

Tasks

1. Work on the GSoC midterm evaluation
2. Fix several bugs in the Pytorch traning loop train_torch.py:
   • Change the model weight matrix from a torch.tensor to a nn.Parameter to enable gradient flow.
   • Add in the option of choosing args.pert_form that affects the loss function during backprop.
3. Replicate Figure 2A and 2B using the current Pytorch code
4. Discuss a strange numerical discrepancy issue when writing tests for Pytorch ODE solver. In particular, given identical input and parameters, Tensorflow and Pytorch ODE solvers agreed to the error of 10e-8 for the first 100 time steps, but drastically diverged from 100 to 200 time steps. This phenomenon occurs for all inputs and parameters.

Relevant issues:

• change-util-functions-and-models (branch)
• Numerical instability between Tensorflow and Pytorch (#56)

Notes

Check out my personal GSoC notebook, where I keep my observations and comments in one place.